def combine_same_pitch_for_length(self):
"""
Given a list of note segments such as
[{rhythm: 1/32, pitches: C}, {rhythm: 1/32, pitches: C}],
combines them into a single {rhythm: 1/16, pitches: C} because
the note found spans more than a single segment.
"""
last_encountered = ""
sum_rhythm = 0
for note in self.notes:
current_song = self.notes[note]
for n in current_song["notes"]:
curr_note, curr_rhythm = n["pitches"][0][0], n["rhythm"]
# Check if the segment is a series of rests or not
# If so, treat it like a note change for mapping
if n["amplitude"] < current_song["loudness"] * self.REST_THRESH:
curr_note = "REST"
if last_encountered is not "" \
and curr_note is not last_encountered:
# A note change occurred, record the note and summed rhythm
self.note_list.append((last_encountered, sum_rhythm))
sum_rhythm = curr_rhythm
else:
if not sum_rhythm:
sum_rhythm = curr_rhythm
else:
sum_rhythm = utils.round_note_length_base2(
sum_rhythm + curr_rhythm)
last_encountered = curr_note
self.note_list.append((last_encountered, sum_rhythm))
def find_pitch_sequences(self):
"""
Populates the learner with information about the individual segments
of the song, including the most likely identified pitch and how long
the segment lasted for in terms of note-length.
"""
for song in self.songs:
bpm = song["bpm"]["value"]
data = song["data"]
time = song["time"]["value"]
temp_song_segs = []
# Modularize all notes to be in the key of C,
# we will use this to shift notes later.
key = data.analysis.key["value"]
for seg in data.analysis.beats:
rhythm = utils.round_note_length_base2(
float(seg.duration) / bpm * SECONDS_PER_MIN)
likely_pitches = sorted(enumerate(seg.mean_pitches()),
key=lambda x: x[1])
pitches = []
for a in likely_pitches:
transposed = a[0] - key
if transposed < 0:
transposed = 12 + transposed
pitches.append((self.PITCH_DICTIONARY[transposed], a[1]))
temp_song_segs.append({"count": seg.absolute_context()[0],
"rhythm": rhythm, "pitches": pitches,
"amplitude": seg.mean_loudness()})
self.notes[song["file"]] = {
"notes": temp_song_segs,
"loudness": data.analysis.loudness
}